Reproduce “Against All Likelihood” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
David Lopez-Paz
The authors propose a language-model procedure that tests whether an answer remains stable across relevant changes in context. The model generates perspectives and answer proposals, then retains their common invariant content, unlike self-consistency, which varies samples under one prompt. A regression example illustrates the idea. The paper offers no empirical evaluation of truth-detection accuracy; agreement can preserve shared errors.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.